We propose to develop new methods of determining the structural and dynamic implications of nuclear magnetic resonance data on proteins in solution. These methods of analysis will contribute in an essential way to our understanding of protein structure and function, and hence to the solution of many important problems in biology and medicine. The analysis itself consists of three phases: In the first, a large ensemble of protein conformations that are geometrically consistent with the data is computed. In the second, these are refined so that they are also consistent with the established physical and chemical regularities of protein structures. In the third, we attempt to find relative populations and models for the fluctuations in these conformations that enable us to back-calculate the NMR data from the ensemble as a whole.